GOMO Example

• Description
• Model Architecture
• Dataset
• Environment Requirements
• Quick Start
• Script Description
  • Script and Sample Code
  • Training Process
• Model Description
  • Evaluation Performance
• Description of Random Situation
• ModelZoo Homepage

Description

Generalized Operator Modelling of the Ocean (GOMO) is a three-dimensional ocean model based on OpenArray which is a simple operator library for the decoupling of ocean modelling and parallel computing (Xiaomeng Huang et al, 2019). GOMO is a numerical solution model using finite differential algorithm to solve PDE equations. With MindSpore and GPU, we can achieve great improvements in solving those PDE equations compared with CPU. This is an example of training GOMO Model with MindSpore on GPU.

Model Architecture

The overall model architecture of GOMO is show below:link. The fundamental equations and algorithms of GOMO can also be found in this article

Dataset

Dataset used: Seamount

• Dataset size: 65x49x21

• Data format：nc

• Download the dataset

download the GOMO(https://github.com/hxmhuang/GOMO) from Github and you can find the seamount dataset file in the GOMO/bin/data directory.

Environment Requirements

• Hardware: GPU
  • Prepare hardware environment with GPU processor.
• Framework
  • MindSpore
• For more information, please check the resources below:
  • MindSpore Tutorials
  • MindSpore Python API

Quick Start

After installing MindSpore via the official website, you can start training as follows:

# run distributed training example
bash run_distribute_train.sh [im] [jm] [kb] [step] [DATASET_PATH]

Script Description

Script and Sample Code

└── ocean_model
    ├── README.md                                 # descriptions about ocean model GOMO
    ├── scripts
    │    ├── run_distribute_train.sh               # launch distributed training for GPU
    ├──src
    │    ├── GOMO.py                               # GOMO model
    │    ├── Grid.py                               # grid initial
    │    ├── stencil.py                            # averaging and differential stencil oprator
    │    ├── op_operator.py                        # averaging and differential kernel operator
    │    ├── read_var.py                           # read variables from nc file
    ├── train.py                                  # train script

Training Process

bash run_distribute_train.sh [im] [jm] [kb] [step] [DATASET_PATH]

Training result will be stored in the current path, whose folder name begins with "train".

Model Description

Evaluation Performance

Description of Random Situation

ModelZoo HomePage

Please check the official homepage.